Workshop Proceedings: Short Papers

Potential for the use of Remotely Operated Vehicles (ROVs) as a platform for passive acoustics

Rodney Rountree¹, Francis Juanes², and Joseph E. Blue^3
1Program Manager, Mount Hope Bay Natural Laboratory, School for Marine Science and Technology, UMASS Dartmouth, 706 Rodney French Blvd., New Bedford, MA 02744-1221 rrountree@UMassD.Edu
²Department of Natural Resources, University of Massachusetts, Amherst, MA 01003
³President, Leviathan Legacy, Inc., 3313 Northglen Drive, Orlando, FL 32806 jblue46498@aol.com

Introduction
We are still largely ignorant of the distribution and behavior of the great majority of marine fish. Possibly one of the greatest challenges to researchers attempting to study the behavioral ecology of fishes is that of finding the fish in the first place. Since some fish are soniferous, acoustic detection and tracking may offer methods of population assessment for management decisions. Passive acoustic techniques can be a valuable tool for the identification of essential fish habitats (EFH) for soniferous species. These techniques can allow for non-destructive surveys of large areas to pinpoint habitats frequented by soniferous species, particularly during spawning events when vocal activity tends to be greatest. Studies of fish sounds can provide a wealth of data on temporal and spatial distribution patterns, habitat use, and spawning, feeding, and predator avoidance behaviors. Currently most investigators use simple omnidirectional hydrophones and can usually only locate the general area of a spawning aggregation, but have often been forced to use circumstantial evidence of the identity and behavior of the calling species (e.g. Saucer and Baltz 1993, Luczkovich et al. 1999a,b). Attempts to use passive acoustics as a tool to identify EFH based on spatial patterns in sound production is also critically hampered by the lack of sufficient data describing the sound characteristics of individual species and behaviors under field conditions. We propose that acoustic technologies utilizing hydrophone arrays to home in on sound sources can greatly improve the study of soniferous fishes and their habitat requirements. First, homing in on sound sources will provide a valuable new tool to validate the identity of sound producers, especially when coupled with underwater photography or video devices. Second, the ability to home in on vocal fishes would enhance our ability to correlate fish sound production with specific locations and habitats. In this paper, we describe our preliminary attempts to develop a Soniferous Fish Locator (SFL) for use with a remotely operated vehicle (ROV) to home in on fish sound sources and make recommendations for future efforts.

Tracking and Homing Basics
The use of passive acoustics for homing was developed for naval warfare during and before WW II. A passive acoustics homing system was implemented on torpedoes for destruction of ships and submarines. The technique of homing was extended to detection and tracking of submarines by sonobuoys during World II. These systems were developed before the advent of small, fast computers and were implemented with electronics that are now known as operational amplifiers. Adequate Signal-to-Noise ratios were required for implementing these techniques. Homing on ships and submarines by torpedoes requires only 2 directional hydrophones because the torpedo body blocks out sound from behind it. The available aperture is small so frequencies, such that there are several wavelengths across the directional hydrophones, are used for the lowest frequency in the tracking bandwidth. The torpedo determines the bearing from acoustic signature (signal) of the ship or submarine by cross-correlating the signatures from the 2 hydrophones. The cross-correlation function is:

(1)

where s₁(t) and s₂(t) are the noise-free time signals from the 2 hydrophones, t is the time delay between the arrivals of the signals at their respective hydrophones and T is the period over which the cross-correlation is estimated. The longer T is, the better the cross-correlation estimate E[C₁₂(t )]. For this function, there are 2 bearings or values of t that can arise for the maximum value of E[C₁₂(t )]. One represents the back direction that we know is wrong because the body of the torpedo blocks out that back direction. The other bearing then has to be the correct one. Hydrophone separation, x, in homing torpedoes is small but a relatively broad segment of the noise spectra is available to provide sufficient bearing accuracy for tracking.

The cross-correlation function of Equation (1) cannot be realized in practice but only estimated. Accuracy of the estimate depends primarily on 1) signal-to-noise ratio (SNR), 2) bandwidth of the signal and 3) separation of the hydrophones. Obviously for wider separations, bearing accuracy is better. The choice of hydrophone separation is a compromise imposed by the operational requirements arising when one wishes to place hydrophones on an ROV. Loss of coherence depends on environmental conditions and their effect on propagation of sound. Loss of coherence is more severe at the higher frequencies, but it is not an important factor for arrays that will fit on an ROV. For large signal bandwidths, the peak of the cross-correlation function is narrow. When estimating the cross-correlation function, the time gate T imposes a (sin pf)/pf type function upon the estimate that, along with SNR, determines how accurately one can track a fish. When the bandwidth is large and SNR high, one can choose a small time window and/or a small hydrophone separation and get good homing results.

Methods

Figure 1. Illustration of the principle of null steering on an acoustic source with two cardiod hydrophones. The Soniferous Fish Locator consists of three hydrophones (H1-H3) configured to form two orthogonal cardiods shown by the solid and dotted lines. The two cardiods are 180 degrees out of phase with each other. Summing the two results in a null along the x-axis. A bearing to a sound source is obtained by rotating the SFL until the source direction is coincident with the null.

The SFL was designed to work based on the well-understood principle of null steering on an acoustic source with two cardiod hydrophones (Fig. 1). Specifically, the SFL consists of three hydrophones configured to form two orthogonal cardiods shown by the solid and dotted lines (Fig. 1). The two cardiods are 180 degrees out of phase with each other in this configuration. Electronic summing of the two cardiods results in a null along the x-axis (Fig. 1). A bearing to a sound source is obtained by rotating the SFL until the sound direction is coincident with the null. To enable an operator to determine bearing, output from the SFL will be sent to both earphones and a recording device. The null is found by listening to the summed output of the two cardiods in one ear while simultaneously listening to the sound intensity with the other ear. Feeding output from cardiod 1 prior to summation with cardiod 2 to the second earphone channel eliminates noise from behind the SFL (sound from behind the SFL is nulled out by the cardiod, Fig. 1). Hence, the operator listens only to the intensity of sounds coming from in front of the SFL. This is an important property of the SFL that reduces interference due to ROV noise and/or boat noise when operating in shallow water. The distances, d, separating the hydrophones can be changed to increase or decrease the sensitivity of the dipole and allow the operator to tune the maximum sensitivity toward the predominant frequency band of the type of soniferous fish species for which he/she is searching. s

Initial tests of the feasibility of deploying an array of hydrophones on a Phantom III model ROV as part of the SFL were conducted in test tanks located at the Northeast-Great Lakes Center for the National Under Sea Research Program at Avery Pt. Connecticut in October 2001. Test were conducted on the array configuration, attachment methods and ROV noise production. The ROV was not able to support a hydrophone array in the required configuration (Fig. 1) because of ballast problems. We therefore had to modify the configuration so that the hydrophone array could be supported by the ROV frame (Fig. 2). Unfortunately, this configuration does not allow for the cancellation of ROV noise (the array must be forward of the noise source as in Fig. 1). With this configuration, noise levels under various operating conditions were tested: 1) with all thrusters off and the ROV sitting on the bottom, 2) with top thrusters on, 3) with rear thrusters on, and 4) with all thrusters on.

Figure 2. Schematic illustration of the hydrophone array configuration and attachment to the Phantom III ROV.

Field testing was conducted within the Stellwagen National Marine Sanctuary on board the R/V Connecticut from October 17-24, 2001. Ten ROV dives were conducted in sand, gravel and boulder habitats within the sanctuary. Operations were conducted in depths of up to 70 m under sometimes harsh environmental conditions and strong tidal currents. To reduce ship noise, ROV dives were conducted while the ship was at anchor and running off of its generators. The array was composed of three TH608-40 model hydrophones made by Engineering Acoustics, Inc (933 Lewis Drive, Suite C, Winter Park, FL 32789). The hydrophones had a nominal sensitivity at the preamplifier output of –160.5 dB. The 3-channel audio data from the array was captured to a laptop PC with a 4-channel I/O board and NIDisk software supplied by Engineering Design (43 Newton St., Belmont, MA 02478). Sound signal processing was conducted using Signal 4.0 (Engineering Design). A 1 k Hz sine wave was played through a portable CD player into the system and the input voltage recorded at the beginning of each ROV dive. This allowed calibration of the system gain, in addition to the hydrophone. A single channel of audio data was simultaneously recorded to video (both Hi-8 and VHS) for backup. The calibration signal was also recorded to the videotape so that calibrated audio data can be obtained directly from the tapes to obtain signal source levels.

Results
Tank tests revealed a very high level of noise, even with the thrusters turned off and the ROV sitting motionless on the bottom of the tank (Fig. 3). Although noise levels were highly variable, we estimated levels of >130 dBV with thruster off and >160 dBV with all thrusters on. The high level of noise precluded the operation of the SFL with a "flying" ROV, with the current array configuration. We therefore decided to modify the operation of the ROV while in the field in order to increase the signal to noise ratio enough to obtain bearing information. We required the ROV to remain stationary with its thrusters turned off long enough to acquire the bearing to the sound source.

Figure 3. Tank test of noise generation by the ROV with all thrusters off (lower left), top thrusters on (lower right), back thrusters on (upper left) and all thrusters on (upper right). Digitized at 20 k Hz.

With all thrusters on, the ROV produced high levels of sound at both high and low frequencies (Fig. 4). Dominant frequencies were centered on 7-8 kHz. While the stationary ROV was significantly quieter, it still generated substantial noise centered on 8 kHz (Fig. 5). The low frequency noise in Fig. 5 is an artifact resulting from mechanical banging, rubbing and tapping on the tank sides by technicians testing sound reception.

Figure 4. Noise generation from the ROV with all thrusters on. Recorded at 20 kHz.

Figure 5. Noise generated by the ROV while sitting on the tank bottom with all thrusters off.

Recording fish sounds in the field with an array attached to the ROV proved to be very difficult in practice. Strong currents limited our ability to remain stationary on the bottom. The ROV was rarely able to maintain its position on the bottom for more than a few minutes before the operator was forced to turn on its thrusters to stabilize the vehicle. This also required the operator to turn on the ROV lights, thus further disturbing the fishes. Fish sounds were recorded on only one occasion when we were able to maintain the ROV on the bottom with its thrusters and lights off (Fig. 6). A prolonged series of low thumps and growls from a single fish were recorded over a 20 minute period when the ROV was sitting stationary with its lights off. During this time a large cusk, Brosme brosme, was frequently observed hanging around the ROV. It is highly likely that the cusk is the source of the recorded sounds. We estimated the ambient noise (ROV + ship + seas) level at around 134 dBV and the cusk call at around 140 dBV. At other times when the lights and thrusters were on, cusk were only observed in a highly agitated state, and appeared to strongly avoid the ROV.

Discussion
Based on preliminary analysis of these data we feel that the concept of a Soniferous Locator Device is viable. However, current ROV designs preclude optimal configuration of the hydrophone array, requiring the SFL to be operated in a stationary mode. We propose that a vehicle specifically designed for low noise production and capable of carrying an SFL with a 2-3 m base line in its nose would provide an exciting new passive acoustic tool for soniferous fish surveys. The low calling rate of the fish recorded in this study demonstrates that it would be difficult to track fish using the manual null steering method proposed. Faster digital tracking using this same principal would correct this problem and should be implemented in future efforts. However, it is important to point out that data collected during this cruise demonstrates that an ROV can serve as an adequate vehicle for the collection of underwater acoustic data even without the SFL. The ROV with a hydrophone attached would be used to locate an optimum location and then would be set down on the bottom to record sounds. In this way, a roving survey could be conducted.

Figure 6. Recording of ROV/Ship and ambient noise (bottom panel) together with the call of the cusk, Brosme brosme. The spectrum of a 95 second sequence of multiple fish calls of a single fish is shown in the bottom panel. The middle panels contain relative amplitude waveform, spectra and power spectra for a 5 second sequence containing only noise, while the upper panels contain a single fish call (sampled at 20 kHz and filtered above 1400 Hz).

Although cusk have long been considered to be soniferous because of the presence of a sonic muscle, they had never been recorded until Norwegian scientists recently recorded their spawning sounds (Aud Vold Soldal, Institute of Marine Research, Norway, pers. Comm.). The calls apparently resemble haddock spawning calls and are very different from those we recorded during this study. Our recordings were conducted well outside of the spawning season for cusk, so the sounds were likely associated with other behavior (feeding or territorial display). Observations made subsequent to this study revealed that cusk vigorously guard the chum bag attached to the ROV and frequency chase away other fishes, suggesting the species is highly aggressive and territorial. Because so little is known of the cusk's behavior, ecology and habitat requirements, and because it appears to respond well to a stationary ROV with its lights turned off, it makes a promising field study animal for passive acoustic.

A secondary outcome of the cruise was that we obtained sufficient video data to suggest that the behavior of some species is strongly influenced by the ROV and/or the ROV lights. Adult cunner, redfish and pollock obviously avoided the ROV during the day, but pollock were strongly attracted to the ROV at night due to our use of chum and bright lights. The chum attracted swarms of amphipods that in turn attracted a large aggregation of pollock as well as haddock, cod and skates. Cusk were only observed in boulder habitat and avoided the mobile ROV both during the day and night when the lights were on. When only infrared lights were used, the cusk was clearly attracted to the chum bag on the ROV and showed no avoidance of a stationary ROV. Contrastingly, species such as cunner, redfish and silver hake appear to avoid the ROV regardless of whether its lights are on or off, or whether it is moving or stationary. The response of the cusk to the mobile ROV with its lights turned on suggest the species strongly avoids the ROV. It could not be determined whether the lights or the ROV noise caused this avoidance, however subsequent observations of cusk behavior indicate no avoidance of stationary cameras with white lights. We suggest then, that the noise generated by the ROV can be a significant source of bias in studies using ROVs for fish census.

Acknowledgements
We thank Meghan Hendry-Brogan for diligent work in both the field and laboratory to collect and process fish sound data. Rebecca Jordan and David Howe assisted in the field. This project received major funding from the Northeast and Great Lakes National Undersea Research Center, which also provided extensive logistical support. The Woods Hole Sea Grant College Program also provided supporting funds. The Sounds Conservancy, Quebec-Labrador Foundation/Atlantic Center for the Environment provided a stipend for Megan’s fieldwork.

Literature Cited
Luczkovich, J.J., H.J. Daniel, III., M.W. Sprague, S.E. Johnson, R.C. Pullinger, T. Jenkins, and M. Hutchinson. 1999. Characterization of critical spawning habitats of weakfish, spotted seatrout and red drum in Pamlico Sound using hydrophone surveys. Final Report and Annual Performance Report Grant F-62-1 and F-62-2, Funded by the U.S. Department of the Interior, Fish and Wildlife Service in Cooperation with the North Carolina Department of Environment and Natural Resources, Division of Marine Fisheries, Morehead City, NC 28557. 128 p.

Luczkovich, J.J., M.W. Sprague, S.E. Johnson, and R. C. Pullinger. 1999. Delimiting spawning areas of weakfish Cynoscion regalis (Family Sciaenidae) in Pamlico Sound, North Carolina using passive hydroacoustic surveys. Bioacoustics 10:143-160.

Saucier, M.H., and D.M. Baltz. 1993. Spawning site selection by spotted seatrout, Cynoscion nebulosus, and black drum, Pogonias cromis, in Louisiana. Env. Biol. Fish. 36:257-272.

Return to Top | Workshop Proceedings: Short Papers